Science topic
Computer Vision - Science topic
Computer Vision is a for discussion on techniques for aqcuiring and analysing images and other high dimensional data in order to produce information.
Questions related to Computer Vision
I am trying to implement a VQA model in ecommerce, and would love to have a dataset that focuses on fashion (or any ecommerce type of goods). If there isn't an available one, is synthetically generating q&a pairs for a given image a good idea? If so, any idea how to approach such problem ?
Hello,
I'm a masters degree student and I am struggling to find a good thesis topic for my masters degree. I would really appreciate if you can help me.
As you know, biosystem engineering is a major where I can work on both mechanical engineering side of things and electrical/computer engineering side of things. Personally, I am interested in precision agriculture(electrical/computer side) and have academical experience on implementing computer vision models(Generally Deep Learning), analyzing and modeling big data(Generally Machine Learning) and deploying IOT applications.
Thank you for your time.
looking for someone working in psychology to discuss with him or her about my research idea.
Specially the Attractiveness.
Most often training in computer vision task are usually in 2D or 3D. Why can't the images be presented into 1D image to access how DL models will perform?
Is there any AI-related (mainly NLP, Computer Vision, Reinforcement Learning based) journal where I can submit short papers? It should be non-open access.
In which application of Machine Learning ( NLP, Computer Vision, etc ) would we find maximum value with Semi-Supervised Learning and Self-Training ?
In recent years, advancement of AI technologies brought a wide spectrum of applications in Biomedical and Pharmaceutical industry. From Computer Vision enabled by Deep Learning algorithms to AI aided Drug Discovery and Drug Repurposing. What's your opinion on the impacts of AI on Healthcare industry in the future?
I am working on the development of a simulator to generate synthetic data for low-level vision tasks. All you need to know is that a simulator moves multiple foregrounds over a background. I am using Bezier curves as trajectories with randomly picked control points. However, I would like to add more complex trajectories than just a smooth Bezier curve. For example, foreground stopping and changing direction, accelerating and decelerating. Can you think of more examples of physically realistic trajectories and how to create them, please let me know.
My budget is tight and work is computer vision task.
Hello,
I am a young researcher and preparing myself for master's related to image processing, computer vision and ai for autonomous vehicle.
I have worked on various projects but confused with their publication that which one will be better in conference paper, journal or the book chapter. It will be very supportive if anyone suggest the best one?
Colors are weak features from computer vision. But, the colors are widely used in video-surveillance and search of objects. However, currently there are pieces of clothing and vehicles that change color automatically or manually, for example, cars. What will be the future of colors as features in machine learning?
GSD is commonly measured for aerial surveys and influenced by many factors. I am interested to measure the GSD of natural images (say an image of a building). My goal is to measure the size of an object in an image based on GSD.
I would appreciate your help.
I want to work on livestock monitoring using computer vision. Is there any datasets (video recordings) related to this work? and suggest any computer vision applications related to this domain.
I have two geometries (Say A and B) with known x,y,z coordinate. These geometries are similar but not of exact shape and size.
Let me for now consider them in 2D for understanding. A is supposed to be a square of size 1mm by 1mm. B is also a square of size 0.5mm by 0.5mm. However, one edge is distorted (see attached figure). Now I wish to correct it, which means, change it to a square of size 0.5mm by 0.5mm. Is there a way to do it?
I know that we can do feature matching across a temporal sequence of images to check which features belong to the same object.
So, for example, in the first image, I can create a bounding box for an object, and then I can run SIFT on that image and get the features inside the bounding box for that image. Then, I can check which features in the next image and the ones after that match with the set of features in my first image (the ones inside the bounding box) and get bounding boxes for the object in the subsequent frames.
However, I was wondering if there is a deep learning-based method to achieve the same result in an arguably better way.
Transformer models, originally developed for natural language processing, now seem to be growing in popularity for computer vision (e.g., Vision Transformers and variants). I am curious whether researchers/engineers in the deep learning and computer vision communities believe that Vision Transformers will eventually dominate over convolutional models, or not? I am referring to actual deployments, in particular in mobile, embedded and low-power applications.
Recently, MLP-based models such as MLP-Mixer, ResMLP, and ConvMixer have suddenly appeared as a potential alternative to Transformers. MLP-based models are much simpler than Transformers and seem to become competitive in accuracy. A similar question as above arises: do you believe MLP-based models for vision will overtake Vision Transformers and convolutional models, in actual deployments? What's your opinion, and why?
EfficientDet outputs classes and bounding boxes. My question is about both but specifically I am interested in the class prediction net part. In the paper's diagram it shows 2 conv layers. I don't understand the code and how it works. And what's the difference between the 2 conv layers of classification and box prediction?
Hello Researchers,
Can you guys tell me the problems or limitations of Computer Vision in this era, on which no one has yet paid heed or problems on which researchers and Industries are working but still didn't get success?
Thanks in Advance!
The Computer Vision a field of artificial intelligence that trains computers to interpret and understand the visual world. Using digital images from cameras and videos and deep learning models, machines can accurately identify and classify objects.
I am new to arXiv and want to submit a pre-print at arXiv. In order to submit my paper I require three endorsements, of existing arXiv users. I want some to endorse me in Computer Vision and Pattern Recognition category by visiting the below link or using the endorsement code:
Click to Endorse: https://arxiv.org/auth/endorse?x=DYY3VC
or visit the link and enter the code: http://arxiv.org/auth/endorse.php
Endorsement Code: DYY3VC
I have a few peer-reviewed publications in the area of endorsement which can be checked from my profile. Thanks in advance.
It is appreciated if you recommend a benchmark study with the available dataset for video processing.
Lane detection is a common use case in Computer Vision. Self-driving cars rely heavily on seamless lane detection. I attempted a road lane detection inspired use case, using computer vision to detect railway track lines. I am encountering a problem here. In the case of road lane detection, the colour difference between road (black) and lane lines (yellow/ white) makes edge detection and thus lane detection fairly easy. Meanwhile, in railway track line detection, no such clear threshold for edge detection exists and the output is as in the second image. Thus making the detection of track lines unclear with noise from the track slab detections etc.
This question, therefore, seeks guidance/ advice/ Knowledge exchange to solve this problem.
Any feedback on the approach taken to attempt the problem is highly appreciated.
Tech: OpenCV
When training a CNN for object detection within images, it is supplied with images containing examples of said object (or list of objects) through their coordinates within the image.
But what happens if you don't supply the coordinates of all the objects in the training images? So there are some object in your training images that match the pattern that the CNN is looking for but are marked as negatives...
I suspect this has a mostly negative impact but, Are there any resources/papers that go over this issue into details?
Thanks
During preprocessing medical image data different techniques should be considered such as cropping, filtering, masking, augmentation. My query is, which techniques are frequently applied to medical image datasets during pre-processing?
I'm working in a RGB image classification algorythm for blueberry ripeness stages identification in which I must collect photos for the algorythm traning data set. Is it valid taking the photos using a smartphone camera with enought resolution (eg.: Xiami Redmi Note 10, etc.), considering this research is intended to be published on a scientific journal? I've read some papers refering usage of GoPro cameras for collecting image-data sets. I have compared the image quallity between my smartphone and GoPro pics and the first one are quite better.
Thanks for the answers!
Hi,
I am researching about long-range stereo vision systems because I am planning to estimate the depth of an object that is +300m away.
I use the method proposed by Zhang in 2000 to calibrate the stereo system for close-range applications, but I got stuck here. For the described situation, I would need a large chessboard printing, which is almost impossible.
Therefore, I started to read some papers regarding self-calibration algorithms, but I didn't find anything applied to long-range situations like mine.
Firstly, do you think that this task is possible? What would be the limitations?
I am planning to work with a small baseline due to my hardware configuration.
I'll be glad if you provide/recommend any information source.
Thanks in advance.
SVM can be added as the last layer of the CNN model. How to improve the performance of CNN-based SVM?
Hello,
I have been working on computer vision. I used datasets from Kaggle or other sites for my projects. But now I want to do lane departure warning, and real-time lane detection with real-time conditions(illuminations, road conditions, traffic, etc.). Then the idea to use simulators comes to my mind but there are lots of simulators on online but I'm confused about which one would be suitable for my work!
It would be very supportive if anyone guide me through picking up the best simulator for my works.
Is it because the imaging equation used by the color constancy model is built on RAW images? Or is it because the diagonal model can only be applied to RAW images? When we train a color constancy model using sRGB images, can we still use certain traditional color constancy models such as gamut mapping, correction moments, or CNN?
Hello,
We are planning to undertake a clinical study that is centered around recording movement data from IMU sensors alongside video recordings of individuals. We aim to establish a ML model, which can predict mediation states based on motion of individuals. For this purpose, we would like to rate the motion/symptom profile of individuals every, say, 2 minutes on the video. We are looking for a platform, which can help us streamline this process. Ideally, it would display video and IMU data synchronised on the left, and on the right would allow a rater to enter labels in prespecified categories. Having spent some time on google, I can only find computer vision tools, which are not really what we are looking for.
Any suggestions are welcome and would help us out a lot. I am happy to share some more details on what exactly we are looking for :)
Could anyone suggest a software or code (R or Python) that is capable of recognizing bumblebees (recognizing only not identifying) from video recordings?
As part of my coursework, we were required to choose a research subject for our end-of-semester project. I chose to investigate how artificial intelligence might be used to identify and predict the presence of tuberculosis in x-ray images. I am looking for research ideas and materials. I'd be really thankful if anyone could offer a suggestion on how to proceed.
For most drivers who manually control a vehicle, obeying traffic regulations like speed limits looks like a huge task for them as some overlook the importance of these strategically placed limits. Our quest is to devise ways to use AI-based technology in both autonomous and manually controlled cars to detect these speed limits on the road and for the AI technology to proceed to act on it in bringing the vehicle to a lower speed when the vehicle is travelling at a speed out of range. Issuing warning about a speed out of range is another thing and it all falls on the driver to heed to the warnings or not.
I want to research machine learning or computer vision, image processing..Can anyone help?
Hi, I'm a beginner student looking for someone who is interested and passionate about the various subfields in AI, mostly computer vision. I'm really interested in deep learning and want to write a good research paper in computer vision (vehicle detection ) but don't have any friends with whom I can discuss and research on certain topics. I don't know if it's the right place to discuss this but please let me know if anyone is willing to work on research papers together as a team. Thanks
Hello, anyone interested in doing research collaborations? I have done and published some papers regarding deep learning, computer vision, IoT, and security.
Here is my page:
This is also my recent publication in Elsevier.
I need to know do I need to calculate the above metrics by selecting every available trained weight with all the test data to evaluate the model performance? Or do I only need to calculate those metrics for the best weight file I was chosen as my final model for the complete test set? The reason for asking this is in the below reference the author of the repository calculates both metrics by taking every weight file for the complete test set.
Please refer line 152 .
I need to know whether I should consider the number of labels available or the number of images available to increase the model performance. I mean in my case in a single image I have multiple annotated labels available. I know when I increase the number of images for training the accuracy of the model will go high. But in my case do I need to pay attention to maximizing the number of labels in the dataset or the number of images in the dataset to get better accuracy.
Ex:
Case 1:
1 image contains 4 annotated labels(2 for ClassA , 2 for ClassB)
|
|
Total: 60 images
Case 2:
image_b_1 contains 1 annotated label(1 for ClassA)
image_b_2 contains 1 annotated label(1 for ClassB)
|
|
Total: 200 images
Which case will give the maximum accuracy results during the training?
I need to calculate the accuracy, precision, recall, specificity, and F1 score for my Mask-RCNN model. Hence I hope to calculate the confusion matrix to the whole dataset first to get the TP, FP, TN, FN values. But I noticed almost all the available solutions out there for the calculation of confusion matrix, only outputs the TP, FP, and FN values. So how can I calculate metrics like accuracy and specificity which includes TN as a parameter?.Should I consider the TN as 0 during the calculations ?
Hello dear researchers.
I run the siamfc ++ algorithm. It showed better results than the other algorithms but its accuracy is still not acceptable. One of my ideas is to run an object detection algorithm first and then draw a bunding box around the object I want to track.I think by doing this I have indirectly labeled the selected object. So the results and accuracy should be improved compared to the previous case. Do you think this is correct? Is there a better idea to improve the algorithm?
Thanks for your tips
Dear All,
I have to make my own custom data set in TUM dataset format (like https://vision.in.tum.de/data/datasets/rgbd-dataset/file_formats) with ground truth data in order to check the accuracy of SLAM algorithm for Visual - Inertial Scenario.
Any suggestions and recommendations are highly appreciated.
Thank you.
Which object tracking algorithms are highly accurate and also have high processing speed?
Opencv algorithms themselves do not work well at high speeds
Thank you for your help
Improving my lab performance by digital transformation I want to develop a software which perform a calibration of an analog instrument. Advances in computer vision with Pyton addresses the stuf but fot me it has the error that the reading of the gauge depends on angle of the needle, not properly over the scale. So if the scale is not correctly printed that bias will not be detected by the vision. ¿Do you know any alternative?
I have an image from a calibrated camera with all intrinsic and extrinsic parameters. The image contains a table with known corner coordinates (image coordinate system). Now I want to know the projection of the corners on the ground plane. The real-time height of the table is known and I know that the projection will be vertically below the actual corner in the image (unless there is rotation) but the question is How much I should shift it vertically downward?
Refer the attached image.
Ayn Ideas about topics in image processing and computer vision able to be master topics related to recognition and infrared waves and it is great if it related to optical physics field ?!
Im looking for a dataset with images of people from at least four angles and the ground truth anthropometric measurements but am not able to find one. Caesar dataset might be exactly what I need but is not free.
If not, whats the best dataset available for human multi view images without any ground truth? Preferably the subjects should either have their arms raised or have them at a distance from the center part of their bodies.
I am writing a review article on facial analysis in biomedical data science. As part of this there is a section on automatic and manual facial landmarking.
There is a large literature on this topic in computer science, but it is usually more focussed on computer vision than medical applications.
I am wondering if there exists established and respected software for automatic landmarking of facial images in a biomedical context? Any help is much appreciated.
Harry
Are there standard computer vision methods or deep learning architectures where the input is a video of frames and the output is a collection of distinct images extracted from the video. Distinct images might be 10-20% overlap between consecutive frames.
Hi Everyone, I'm currently converting video into images where I noticed 85% of the images doesn't contain the object. Is there any algorithm to check whether an image contains an object or not using the objectness score?
Thanks in advance :)
Hi Everyone,
I'm currently practising an object detection model which should detect a car, person, truck, etc. in both day and night time. Now, I have started gathering data for both day and night time. I'm not sure whether to train a separate model for daylight and another model for the night-light or to combine together and train it?
can anyone suggest to me the data distribution for each class at day and night light? I presume it should be a uniform distribution. Please correct me if I'm wrong.
Eg: for person: 700 images at daylight and another 700 images for nightlight
Any suggestion would be helpful.
Thanks in Advance.
I'm developing a computer vision using python. Is it a good idea to use python for image processing in production? or it's better to write my tests in python and for production use c++?
and can I compile python?
I am highly interested to do a PhD in Artificial Intelligence, could any one suggest me which would be the challenging field in the coming future, and any ideas on the research projects related to Computer Vision, Machine learning, and Neural Networks sectors.
I have been looking into self-supervised methods in computer vision. For example which looks at three consecutive frames of video, tasks a network with predicting the third frame, and uses the original as the ground truth / supervisory signal.
This type of pretext task for self supervision is a cross between context-based and semantic label based pretext tasks
Lane line detection in many ways is approached as a subset of the semantic segmentation task.
I am wondering if there is any way to come up with a pretext task that is specific to lane line detection?
I have seen where self supervision is used in the lane fitting task
But this is used after the lane segments have been identified.
I am looking for Imagenet dataset but I can't seem to find any public version available for it. Where can I get the whole dataset?
Hi! I'm trying to train a convolutional neural network (CNN) using Keras for leaves disease classification from images. There are very few plant disease image datasets, so I need to use one of the available TensorFlow datasets for training my model: specifically, two TensorFlow datasets are suitable for this task: 'plant_village' dataset and 'plant_leaves' dataset.
The problem is I don't know how to explore that datasets for see the classes, features, labels... and I don't know how to split them in training, validation and test datasets. I've tried to use the code that is used in TensorFlow docs to explore and manage 'CIFAR10' or 'mnist' datasets, but it doesn't work with the plants image datasets...
Can someone suggest me how to explore and manage 'plant_village' and/or 'plant_leaves' datasets, please?
I've spent a few years outside of school now but looking to go back for a Masters or PhD. I published two papers in undergrad, but my GPA wasn't great due to health problems that have since resolved. I work as a full-stack Software Engineer but have a BS in Food Systems from the University of Minnesota. I'm considering a few different angles and would love advice.
Angle #1: Join an agriculture department as a Masters or PhD student and focus on the computational parts of agriculture. The entry requirements to different schools seems to vary widely.
Angle #2: Start with a Data Science Masters Degree and go from there. A lot of the programs I've looked at so far are very industry focused and also don't go as deep into theory or cutting-edge work as I'd like. A lot of what they cover I've learned on my own so it just seems a way to get a paper proving that I know it. The programs I've seen seem very focused on just moving you up the corporate ladder when I don't really care about that, I just want to work on cool and impactful problems. On the plus side, many seem easier to get into from what I understand.
Angle #3: Attempt to get in a Computer Science Masters or PhD program with a mediocre GPA and hope my research papers and work experience can speak louder than my meh GPA. Would be especially interested in computer vision with the idea of eventually applying it to cutting-edge CV in agriculture (see angle #1).
Any advice? It could be directly on which angle would work best or other things I might not have thought of while I'm considering grad school. I really loved doing research and want to get back into it. I'm considering becoming a professor someday as from my experience of teaching people to code, I realized I also love that. I'm working for the U of Minnesota so get a good discount on classes so I'm planning on taking a grad-level course on Machine Learning this Fall.
Hello,
Is anyone already worked with MR image data set??? If so, Is there any model to remove the motion artifacts in the MR image data set if contains??? What should we do if we have an MR image with motion artifacts??? Please give me your suggestions if it is possible to remove artifacts once the scan is produced.
Thanks in advance,
Dhanunjaya, Mitta
I have gone through quite a few papers in the domain of computer vision, and what I found that pyramid pooling operations are often found to be superior over regular pooling operations. This has also been empirically shown in several papers. Although, I cannot find any proper reasoning behind that. Most of the papers use very primitive terms and reasons. Can someone please explain the reason in details. It would be really helpful if you could explain from scratch.
I need help for the following problem. Suppose that I have two images of a patient‘s body one today and an older one a couple of years ago. The problem is: find the aerias of the skin where moles have changed color or shape. I am a newbie on computer vision so I am asking for quidance from where to start. Are there any algorithms known to solve this problem or ca you suggest some papers or algorithms to start from??
Hello community. Is there any public datasets we can use to identify and classify the Palm Tree leaves diseases, especially for types/species in the North Africa or Middle east Region?
Many thanks in advance.
Digital Image processing based research topic for the final thesis of Masters, It must not be based on machine learning.
I have some questions, anybody can help me with references!
- Which formula is being used for Aerial Triangulation in Pix4D Mapper?
- Which formula is being used for the Computer vision algorithm, bundle block adjustment and stereo matching algorithm in Pix4D Mapper?
Thanks in advance!
Hi all,
I have a computer vision task in hand. Much as it's quite simple in my opinion, I'm very naive in this area thus looking for the simplest and fastest methods.
There's a laser pointer projected on a screen that keeps bouncing around. I need to capture the location and the velocity of the projected dot with respect to some reference point. I would really appreciate it if someone elaborates on the procedure in simple terms.
Regards,
Armin
Hi, I'm a beginner student looking for someone who is interested and passionate about the various subfields in AI, mostly computer vision. I'm really interested in deep learning and want to write a good research paper in computer vision (medical field like disease detection) but don't have any friends with whom I can discuss and research on certain topics. I don't know if it's the right place to discuss this but please let me know if anyone is willing to work on research papers together as a team. Thanks.
While training my GNN (graph neural network) model, the loss is badly fluctuating. Someone had suggested increasing the batch size or decreasing the learning rate, but the results are remaining the same. Can anyone suggest other possible reasons and remedies for solving this issue?
(In the graphs attached below, the x-axis represents the number of samples and the y-axis represents training loss.)
Hi everyone,
Im willing to test several compression methods (conventional :JPEG , WebP, HEVC and AI BASED: mbt2018, cheng2020_anchor, bmshj2018) on computer vision application.
I will use yolov5 for object detection(https://github.com/ultralytics/yolov5) and i will train my model with coco train Dataset (JPEG)then i will use a subset of coco test dataset (with 100) images to test the performence of my model with the different compression methods.
I don't have a lot of experience with computer vision application and this is my first time with yolov5. So i am wondering if yolov5 can support the different type of images or i will face problem with this part and also if my comparsion result will be accurate enough using this method
thank youu
Dear All,
I tested my SLAM algorithm on EuRoC, KITTI and TUM Datasets. But I want to create my own SLAM datasets with Ground Truth. I heard it is possible with blender, But I really don't know how to create my SLAM Dataset with ground truth because I can't find any helpful tutorial on internet to do that? Can someone help me out in this matter.
Your suggestion and recommendation are highly appreciated.
Thank you.
Same question in
Hi!
I have got some images of transparent gel bead. Because it is transparent, I cannot turn it into masked Black&white image using "8-bit->threshold->set measurements (Feret's Diameter)".
1. is there another way to obtain the outline and turn it into a solid black&white image for diameter analysis?
2. Alternatively, is there a macro script/plugin that are designed for such transparent praticle?
3. in some previous images, the gel beads are overlap with each other. How to process the overlap images for the diameter analysis in batch?
Thank you!
the image scale is 543.75 pixel=100um
Dear Friends and Researcher
What do you think about adding artificial intelligence, deep learning, and computer vision to the scopes of "Iranian Journal of Medical Microbiology"?
How do you think these topics contribute to medical microbiology?
This Journal was recently indexed in SCOPUS. Do you think it is important and help other researchers and microbiologists?
Given image points and corresponding 3D points of 7 objects in an image (all objects in their local coordinate system) how can I find camera pose for each of the objects.
Hi,
I have tested my SLAM algorithm on EuRoC, TUM and KITTI Datasets. But I would like to test on my own custom dataset. Moreover, I already rectify how to make my own dataset in TUM dataset format, but I didn't know how to calculate the ground truth of my own data set so that I calculate the accuracy of my SLAM algorithm.
Would kindly provide me any suggestions and recommendation?
Thank you.
I’m working on an image processing project with aim of developing A method to differentiate between healthy and diseased cornea. I have images taken from SHG microscopy for the collagen bundles in the human corne. I have used GLCM and I obtained the correlation, energy, homogeneity and contrast of each image but the results are not good enough to use them as a solid evidence to classify the images. please can anyone suggest for me any other method that can be used to classify collagen bundles image that can be done witthin a short time or a method to enhance my glcm results.
Thank you in advance
As I know CNN requires images most of the times , but my data frame is taking a size of (335,48) ; which is not an image but numerical values and categorical output, how Can I use CNN or deep learning for this situation ? Thank you..
I have tried based on pixel value with a referential ground-truth temperature of some different pixel value, a region of interest (thermal image of a person's forehead ). I am getting near about temperature but not current.
What else I can do to detect temperature from thermal images.
Please help me to get some references.
I am attaching one sample thermal image, where I am trying to find.
When I read papers in computer vision field, I saw many energy functions or objective functions. I can understand the paper from high level, but when I try to implement the paper I stuck on solving the energy function.
I am implementing DynamicFusion: Reconstruction and Tracking of Non-rigid Scenes in Real-Time, published in 2015 CVPR. The energy function is that:
E(Wt,V,Dt,E)=Data(Wt,V,Dt)+λReg(Wt,E)
The paper use Gauss-Newton to solve the parameters. Should I convert the energy function into a linear system like Ax+b? If so, how to combine Data term and Reg term to be Ax+b?
And why papers usually don't give the deviation of the energy function?
Hello everyone,
I intend to do some investigation on applied Computer Vision on Wood Log Detection. For that I would need a proper sized Image Dataset. Ideally, this should be publicly available to give others the chance for comparison.
The only source I found on this topic was the paper "The HAWKwood Database", but the link given to the Database in [1] is offline.
Anyone knows how to get access to that - or similar - datasets?
Any help is appreciated!
Thank you very much!
Kind Regards,
Jonathan
Hello everyone,
I need to annotate the MRI images (with .nii.gz format). Each image includes several slices. What I need is an annotation tool that can annotate the area of interest and propagate it in all slices? (considering that the location of object changes in different slices).
Thank you all in advance.
I would like to get good books for Image processing, Computer vision, AI, Deep learning, machine learning, and remote sensing. Kindly suggest good authors books for the above topics and its links for online ordering. Thanks in Advance.
Hello, I'm currently studying a Deriche filter. I've managed to create python program for derivative smoothing from this article [1]. As I understand, Deriche used one α for smoothing in both directions - X and Y, so this filter seems to me isotrophic. Is it correct way to use two alphas − α_x and α_y to add anistrophy?
Can someone please explain me the relaion between α in Deriche filter and σ in Gaussian filter? In [1] i found an equation picture, but i do not understand the symbol П. Is it π? To my mind, it's not, because in the article they took α = 0.14 and σ = 10. From these values П = 625/19≈3.1887755.
Thank you.
[1]: Deriche, R. (1990). Fast algorithms for low-level vision. IEEE Transactions on Pattern Analysis and Machine Intelligence, 12(1), 78–87. doi:10.1109/34.41386
I've some experience with computational vision, however only the basics of deep learning. I'm searching for indication of courses with a more practical approach. I've seen many different sites trying to seel courses, and I'm looking for the opnion of real people.
Which course would you indicate about the subject and why?
Thank you!
Hi
I’m wondering why do most Image captioning deep learning models usually follow Image encoder-Language decoder architectures ? I understand that there are many different flavor and models for image captioning but they all most all seems to follow the Image-encoder sentance decoder paradigm.
Why not using Object detector to generate a list of image features. Then using this list of features as am input to seq2seq model that will generate sentences ?
Are you aware of an attempt/an experiment or a paper discussed using seq2seq models for image captioning. ? Is there anything wrong with this approach ?
We are doing a project in which we are detecting (using YOLOv4) runway debris using a fixed camera mounted at a pole on the side of the runway. We want to find out the position of the object with respect to the runway surface. Does anyone know about any algorithm or technique that will help us find out the position of the object with respect to runway dimensions?
Can anyone suggest aero engine fault diagnostic model using image processing techniques (computer vision based model)?
COVID-19 disease, caused by the SARS-CoV-2 (Severe acute respiratory syndrome coronavirus 2) virus, was identified in December 2019 in China and defined / declared aa global pandemic by the WHO on 11 March 2020. Artificial Intelligence (AI) is a potentially powerful tool in the fight against the COVID-19 pandemic. AI, ML can be useful to diagnose, predict, and explain (treat) COVID-19 diseases. How can we use AI, Machine learning, Computer Vision, Data Science or computational science to cope with COVID-19? What are the potential AI, ML use cases? and help manage socio-economic impacts. Since the outbreak of the pandemic, there has been a scramble to use and explore AI, and other data analytic tools, for these purposes.
I would like to know more about the following issues:
1. Using images as feedback for the closed-loop control of civil structures
2. Simulating the vision-guided control on the computer without doing an experiment
P.S: There is a wealth of literature that is related to robot control. But, I intend to know more about civil structures, i.e., buildings and bridges.
Hello all,
I want to utilize some method that detects edges from unorganized point clouds. The cloud scenes would contain objects such as buildings, roads, streets, ditches etc. Still, I would like that the method would be as universal as possible.
So far I have studied that good results have been found through eigen value analysis of point neighborhood's covariance matrix. Some also use principal component analysis to process the eigen values further. [
Conference Paper Fast and Robust Edge Extraction in Unorganized Point Clouds
]Another paper uses plane fitting and some threshold parameters for edge detection. Results seem promisin, but there are some parameters to fix on the other hand [https://www.mdpi.com/2072-4292/8/9/710]
I was thinking of trying the eigen analysis first.
Any opinions or other suggestion?
Basically, in the computer vision applications, recently there are more focus on mathematical approach. Deep learning is used in almost all tasks in computer vision. Using geometric algebra etc. how we can improve the state of arts methods in different computer vision tasks.
I am using data augmentation and decaying learning rate after each epoch. If I don't use data augmentation but keeping callback, then my training accuracy reaches to 99.65% but not validation. Another scenario is, if I remove callback of learning rate decay but keeping data augmentation then training accuracy also improves and reaches 99% but not validation. Why is it stucking with current configuration (lr decay + Data Augmentation) ?
What could be the reason to have this problem with data augmentation?
Dear Research Community.
I am new to the field of deep learning based object detection techniques. At this time, I am working on a project which will not use front-facing camera to detect people. Nearly 99% of the datasets that I have across uses some form of front-facing camera image to create its database. For the literature survey I have done, it is suggested that the YOLO model should be given as much relevant data as possible as for DNN based techniques, their accuracy scales with the number of training samples provided.
Thus, if you would suggest some tools and techniques using which I can use to create my own custom dataset, I will appreciate it.
With best regards,
Azmyin
Hi,
I want to apply GAN as a data-augmentation technique in my model. I am bit confused regarding implementation that how could I give the generated images of GAN to my model on run-time? because GAN takes time to generate good and sensible images. At what point I could feed GAN generated Images to my model and what would be the best approach? In whole scenario, I want to skip saving Images on the disk and then preprocessing the images to feed into my model.
Looking forward to your kind suggestions!
what is the best machine learning method , to segment and extract the length , angle and the size of collagen bundles pictures obtained from the human cornea?
Please tell me the best algorithm for face detection and re identification ?
Hi,
I am training my models to classify 10 classes. Some of the classes are easy for my models and achieve accuracy greater than 95% but some classes have very hard patterns and accuracy stops at 83% which sucks overall accuracy of my models. I am struggling to find any way which could deal with hard pattern to improve overall accuracy.
How could I achive better accuracy on hard examples?
Thanks in advance!
+ Tensor algebra can reduce the parameters of CNN models.
+ Tensor algebra can model the causal structure of data formation.
What are the other tensor roles in Computer Graphics, Computer Vision, and Machine Learning?
I have 3 frames. I am trying to calculate the camera pose at third frame by iterative optimization. Using the first two frames, I've calculated 3D points by triangulation. Then, I projected some of these points onto the 3rd frame. Now, I can find the re-projection error but couldn't figure out how to optimize it with sparse solvers.
I'm prototyping with python. I decided to use the g2opy library for the whole project. I've already looked through the examples but there is no detailed information and no documentation. The examples are about full/sparse bundle adjustment and pose graph optimization.
Can you suggest any documentation / example ?
Hello everyone,
I’m working on a project in which I need to detect anomalies in a particular scene (two background scenes). The anomaly could be anything (bolts, pliers, glasses, etc.). However, I have generated synthetic data training with unity because I have very few realistic images and here comes the problem. I was looking throughout different techniques like Domain Adaptation since I need to find a solution that implies training on synthetic and doing inference on real ones, but each of them seems to be focused on the class of the image and thus could fail in finding particular anomaly. I’m not an expert in this field and I’d like to hear an expert’s opinion since I am a little bit stuck :(
thank you very much for you answers!
I want matlab code to apply LBP feature extraction in an Image and save the extracted features in a feature vector...any help?
I have seen that a center loss is beneficial in computer vision, especially in face recognition. I have tried to understand this concept from the following material
- A Discriminative Feature Learning Approach for Deep Face Recognition
- https://www.slideshare.net/JisungDavidKim/center-loss-for-face-recognition
However, I could not understand the concept clearly. If someone can explain with the help of an example, that would be appreciated.
I want my mesh to be as similar to the input image from the given angles as possible. So im taking projection of the mesh and comparing it with my input image. Currently im taking the outline of both the mesh and the object in the input image and applying Mean Square Error to compare them. Is there a better way?
